Many to Many single query optimization - mysql

Question

I have 3 tables with many to many relation
CREATE TABLE news(id int, content varchar(64));
CREATE TABLE tags(id int, name varchar(64));
CREATE TABLE news_tags(id int, tag_id int, news_id int);
INSERT INTO news VALUES
(1, "Hello, world!"),
(2, "Test news"),
(3, "test news 2"),
(4, "test news 3"),
(5, "test news 4");
INSERT INTO tags VALUES
(1, "general tag"),
(2, "sub tag 1"),
(3, "sub tag 2"),
(4, "normal tag");
INSERT INTO news_tags VALUES
(1, 1, 1),
(2, 2, 1),
(3, 3, 1);
INSERT INTO news_tags VALUES
(4, 1, 2),
(5, 2, 2),
(6, 1, 3),
(7, 4, 3),
(7, 2, 4),
(8, 3, 4),
(9, 1, 5);
I want to select news_id what
On relation have only general (id 1 on example) tag and dont have any other sub tag (on exmpl id 3)
Have pair of tags general + sub tag (id 2)
I create a query
SELECT news_id FROM news_tags WHERE tag_id = 1 OR tag_id = 2
GROUP BY news_id
HAVING COUNT(news_id) = 2
UNION
SELECT news_id FROM news_tags WHERE tag_id = 1 AND news_id not in (SELECT news_id FROM news_tags WHERE tag_id in (2,3));
but have 2 problems
I think its not optimization way (have 2 select with union + sub select query)
if i what search more one pairs sub tags i need add new select with union
How can i optimizate this query ?
live example http://www.sqlfiddle.com/#!9/1067b7/1/0

Answer 1

Your question is unclear because the concepts of "sub" tags and "general" tags are not defined.
But if you want to handle multiple conditions at the same time, you can still use one GROUP BY and HAVING clause.
For instance, if you wanted news_ids that met either of these conditions:
tag_id = 1
Or both tag_id = 2 and tag_id = 3
Then you can use:
SELECT nt.news_id
FROM news_tags nt
GROUP BY nt.news_id
HAVING (COUNT(*) = 1 AND MIN(nt.tag_id) = 1) OR
SUM( nt.tag_id IN (2, 3) ) = 2;
You can easily extend this idea to the descriptions of the tags (but you need to join in the tags table for that.

Answer 2

I would suggest based on the extra comments about blocking tags, to do a redesign.
Assigning tags to news items is good, but your table should look like news_tags(news_id, tag_id), and primary key is over both news_id, and tag_id field.
If you want to make tags blocking, one way is to add another Many-To-Many relation, called news_blocking_tags(news_id, tag_id). Or you can define your news_tags(news_id, tag_id, is_blocking), so you know which tags are blocking, and which are just tags.
Optimising starts with designing the database. We can only give general pointers here. Good that you know what the outcome needs to be, that's already half the design!

Answer 3

Look like i found solution without made change on schema tables
SELECT news_id, tag_id
FROM news_tags
WHERE tag_id in (1,2,3)
GROUP BY news_id
HAVING (COUNT(news_id) = 1 AND tag_id = 1) OR (count(news_id) = 2 and tag_id in (1,2));
first we find all news with blocked tags , than with HAVING filter result
(COUNT(news_id) = 1 AND tag_id = 1)
find all record with only "blocked" tag for all countrys (its part can use for on many to many find all records what use only single tag)
(count(news_id) = 2 and tag_id in (1,2))
find pairs of "blocked" tag and country tag
if we need more country we need add OR and
(count(news_id) = 2 and tag_id in (1,3))
i need do some test but look like its work nice , and better that my first query

Answer 4

I've done tagging many times. I recommend not using a many-to-many mapping table to a tag table. Instead, combine them into
CREATE TABLE tags (
news_id MEDIUMINT UNSIGNED NOT NULL, -- Assuming you don't need more than 16M
tag VARCHAR(...) NOT NULL,
PRIMARY KEY(news_id, tag), -- For going one way
INDEX(tag, news_id) -- For going the other way
) ENGINE=InnoDB; -- Important due to the way indexes are handled
Adding an auto_inc is a waste of space and speed.
Yeah, some queries get a big gnarly. But any technique leads to messy code; I belive that this schema is the best.